Hi, I'm Sashank!
I'm a research scientist with 10+ years of experience studying the failure modes of intelligent systems. Currently, I work on post-training & alignment of open frontier models at Reflection ↗, where I drive our reward modeling efforts.
Previously, I lead the research team at atla ↗, training general-purpose evaluators ↗. Before that, I was at limbic ↗ building safe and performant clinical AI ↗. During my postdoc at Princeton ↗ I studied failures of continual learning ↗.
I enjoy making music and writing poems ↗. email: firstname[dot]lastname[at]gmail.com
